The Si(ll3) surface is very stable despite its high index but until now its atomic structure has been uncertain. Using a scanning tunneling microscope, we have obtained images of both empty and filled states which provide strong evidence for a particular structural model with a 3x2 unit cell. We explain our results in terms of a general rehybridization principle which accounts for the low surface energy and the spatial distribution of empty and filled states. Our images reveal a high density of domain boundaries which introduce energy states that pin the Fermi level and explain earlier reports of a 3 x 1 reconstruction.PACS numbers: 61.16.Di, 61.14.Hg, 73.20.AtHigh-index Si surfaces tend to be unstable and to facet into lower-index planes upon annealing. 1 ' 2 However, the Si(113) surface constitutes an interesting exception. 3,4 Gibson, McDonald, and Unterwald have annealed thin specimens of (110>-oriented Si in vacuum and observed primarily facets of (111), (100), (110), and (113) orientation. 3 This indicates that the Si(113) surface has an energy that is comparable to that of the low-index surfaces, and implies that it has potential as a substrate for growth. 4 For epitaxial growth on III-V materials the {113} face has already shown crystalgrowth properties that are superior to those of the lowindex surfaces. 5 The structure of the Si(113) surface is important for two reasons. First, it may give clues to a general understanding of why certain faces of Si as well as other tetrahedrally bonded semiconductors are stable. Second, it is necessary step in understanding growth. Previous low-energy electron-diffraction (LEED) experiments suggest that the surface could have both 3 x 1 (Refs. 6 and 7) and 3x2 (Ref. 7) reconstructions, but the atomic arrangement is uncertain. In this Letter we report on the use of scanning tunneling microscopy (STM) to establish both the atomic and electronic structure of the Si(l 13) surface. This enables us to draw conclusions about the energetics that control the reconstruction not only of this surface but of Si and Ge surfaces in general.The samples were cut from a 2xl0 18 -cm -3 Sb-doped single-crystal ingot and had a surface normal within 0.5° of the (113) direction. The surface was mechanically polished, followed by a Shiraki-type oxide etch-regrowth procedure. In UHV the sample was heated to 870 °C for 1 min to remove the surface oxide layer. This treatment generally resulted in surfaces with a 3 x 2 LEED pattern and atomically flat areas larger than 1000x1000 A 2 as observed by STM. Figure 1 (a) shows an STM image recorded at a constant tunneling current I t of 2 nA and a sample bias voltage V s of -2 V corresponding to filled sample states. Figure 1(b) shows an image acquired with // =2 nA and V s = + 2 V corresponding to empty sample states. The star in both images marks equivalent locations. This registry was unequivocally established, over several samples, both by current imaging tunneling spectroscopy 8 (CITS) and by alignment of defect structures in images of the same are...